[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84673-en":3,"doc-seo-84673-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},84673,4810365810221,"Aurora","https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d",8,"Research & Report","Rank Order N of M Codes for Sparse Distributed Memory","Continual learning with large language models remains constrained by expensive retraining and by retrieval-augmented generation’s reliance on external vector stores that lack online associative-memory behavior. The paper evaluates Rank-order N-of-M codes for Sparse Distributed Memory as an alternative to CALM’s threshold-binary encoding, using faithful reimplementation to verify equivalence and then testing capacity and BER robustness under multi-seed interference noise. Results separate representation effects from MAX-Hebbian learning and quantify geometric stability metrics.","Rank-Order N-of-M Codes for Sparse Distributed Memory: Disentangling Representation and Learning Effects in Noise Robustness Against Contemporary Neuromorphic Architectures  \nJoy Bose  \nIndependent Researcher  \nBengaluru, India  \n[joy.bose@ieee.org](joy.bose@ieee.org)  \nAbstract  \nLarge language models excel at reasoning and generation but remain limited as continual learning systems. Incorporating new knowledge requires expensive retraining, while retrieval-augmented generation relies on external vector stores that are not designed as online associative memories. CALM (Nechesov and Ruponen, 2025) addresses this by combining Sparse Distributed Memory with dual asynchronous transformer modules, using SDM as an explicit episodic memory layer that can write and read patterns online without parameter updates. CALM explicitly identifies its threshold-binary encoder as a limitation, noting that improved encoding schemes are future work. Rank-order N-of-M codes for SDM (Furber et al., 2007) offer a concrete alternative: the top-k active dimensions carry geometrically weighted significance values that preserve relative magnitude information while maintaining sparsity. These codes were developed as part of aneuromorphic sequence machine predating modern transformers and have not been evaluated against contemporary SDM-based systems. This paper provides that evaluation. Three contributions are made. First, a faithful reimplementation confirms that WheelSDM and RankOrderSDM are exactly equivalent (cosine similarity 1.0000, 10 seeds) and that RDLIF neurons diverge under interference as documented in the original papers. Second, multi-seed capacity experiments show RankOrderSDM outperforming StandardSDM at saturation by 13.4ppin the scaled-down configuration (W=256, D=64, 10 seeds, p \u003C 0.001, Cohen's d=4.61) and by 0.8pp at the published architecture default (W=4096, D=256, 5 seeds, p=0.031) . Third, BER robustness experiments disentangle two separable mechanisms: a large combined advantage (+38 to 66pp) from rank-order encoding interacting with MAX-Hebbian learning, and a statistically marginal encoding-only advantage under symmetric perturbation (N=30, p > 0.05 at most noise levels), with Kendall tau (0.847) and Jaccard active-set overlap (0.891) quantifying the geometric stability. As a secondary finding, component-level encoding energy at D=4 precision is 2x lower than SpikingMamba's SI-LIF (Huang et al., 2026), though system-level decoder costs dominate in both architectures. Experiments on GloVe-100 word embeddings confirm the small encoding-only advantage on real structured data (+0.1-0.4pp, significant at 3 of 5 BER levels); sentence embeddings (all-MiniLM-L6-v2, D=384, Wang et al. 2020) show a ceiling effect at T=20 patterns.  \n1. Introduction  \nLarge language models have transformed natural language processing but remain limited as continual learning systems. Incorporating new knowledge requires expensive retraining, while retrieval-augmented generation relies on external vector stores that are not designed as online associative memories. CALM (Nechesov and Ruponen, 2025) addresses this by combining Sparse Distributed Memory with dual asynchronous transformer modules, using SDM as an explicit episodic memory layer that can write and read patterns online without parameter updates. This renewed interest in memory-centric AI raises a concrete engineering question that CALM leaves open: what encoding scheme should map continuous neural activations into the SDM address space?  \nCALM currently uses threshold-binary encoding, binarising each embedding dimension against its mean, and explicitly names semantic hashing and binary autoencoders as future directions. Rank-order N-of-M codes (Furber et al., 2007; Bose et al., 2005) offer a concrete alternative: the top-k most active dimensions carry geometrically weighted significance values, preserving relative magnitude while maintaining sparsity. These codes were developed as part of a neur","cbCaiq8zhs1E54zx","https://ap.wps.com/l/cbCaiq8zhs1E54zx","pdf",569434,2,1,16,"English","en",105,"# Introduction\n# Background\n## Rank-Order N-of-M SDM\n## CALM and threshold-binary encoding","[{\"question\":\"What problem does the paper address in continual learning and memory-augmented generation?\",\"answer\":\"It addresses how to add new knowledge online without costly retraining, and how to perform associative memory operations rather than relying on external vector stores that are not designed for online associative behavior.\"},{\"question\":\"How does the paper evaluate whether the published rank-order SDM architecture reproduces prior claims?\",\"answer\":\"It uses a faithful reimplementation and reports that WheelSDM and RankOrderSDM are exactly equivalent across multiple seeds, while certain neurons diverge under interference as documented in earlier work.\"},{\"question\":\"What mechanisms does the paper disentangle regarding BER robustness improvements?\",\"answer\":\"It separates gains arising from rank-order encoding interacting with MAX-Hebbian learning from smaller or marginal encoding-only effects under symmetric perturbation, while using metrics like Kendall tau and Jaccard overlap to assess geometric stability.\"}]",1784197589,40,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"rank-order-n-of-m-codes-for-sparse-distributed-memory","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,47,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":20},"https://docshare.wps.com/document/","Document",{"item":48,"name":12,"@type":43,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/rank-order-n-of-m-codes-for-sparse-distributed-memory/84673/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-23","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does the paper address in continual learning and memory-augmented generation?","Question",{"text":75,"@type":76},"It addresses how to add new knowledge online without costly retraining, and how to perform associative memory operations rather than relying on external vector stores that are not designed for online associative behavior.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the paper evaluate whether the published rank-order SDM architecture reproduces prior claims?",{"text":80,"@type":76},"It uses a faithful reimplementation and reports that WheelSDM and RankOrderSDM are exactly equivalent across multiple seeds, while certain neurons diverge under interference as documented in earlier work.",{"name":82,"@type":73,"acceptedAnswer":83},"What mechanisms does the paper disentangle regarding BER robustness improvements?",{"text":84,"@type":76},"It separates gains arising from rank-order encoding interacting with MAX-Hebbian learning from smaller or marginal encoding-only effects under symmetric perturbation, while using metrics like Kendall tau and Jaccard overlap to assess geometric stability.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,119,122,127,130,134],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":29,"slug":118},7,"Healthcare","healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]